Re: slab corruption in skb allocs

[Scott Feldman]

On Mar 6, 2005, at 10:40 AM, Richard Fuchs wrote:

> Scott Feldman wrote:
> > A bug in the driver.  I have a hunch: please try this patch with 
> > 2.6.9 or higher:
> > http://marc.theaimsgroup.com/?l=linux-netdev&m=110726809431611&w=2
>
> bingo, that fixes it. too bad neither this patch nor the removal of 
> the NAPI config option made it into 2.6.11...

Jesse Brandeburg @ Intel found the fix for the bug but I don't think 
it's been pushed out to Jeff's tree yet, AFAIK.  Soon, I would guess.

Would you mind giving this patch a try against 2.6.11?  I think it's 
equivalent to Jesse's patch, but less intrusive to the driver.

--- linux-2.6.11/drivers/net/e100.c.orig        Sun Mar  6 20:58:15 2005
+++ linux-2.6.11/drivers/net/e100.c     Sun Mar  6 21:01:34 2005
@@ -1471,8 +1471,12 @@ static inline int e100_rx_indicate(struc

        /* If data isn't ready, nothing to indicate */
        if(unlikely(!(rfd_status & cb_complete)))
-               return -EAGAIN;
+               return -ENODATA;

+       /* This allows for a fast restart without re-enabling 
interrupts */
+       if(le16_to_cpu(rfd->command) & cb_el)
+               nic->ru_running = 0;
+
        /* Get actual data size */
        actual_size = le16_to_cpu(rfd->actual_size) & 0x3FFF;
        if(unlikely(actual_size > RFD_BUF_LEN - sizeof(struct rfd)))
@@ -1527,7 +1531,11 @@ static inline void e100_rx_clean(struct
                        break; /* Better luck next time (see watchdog) 
*/
        }

-       e100_start_receiver(nic);
+       /* NAPI: attempt to restart the receiver iff the list is
+        * totally clean otherwise we'll race between hardware and
+        * nic->rx_to_clean. */
+       if(!work_done || *work_done == 0)
+               e100_start_receiver(nic);
 }

 static void e100_rx_clean_list(struct nic *nic)


> > No.  e1000 is a totally different driver/device with very similar 
> > name.
>
> too bad, i was hoping for an explanation for some unexplainable 
> crashes i've been experiencing... ;)

Take the e1000 issue to linux-netdev.

-scott

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/